1 / 44

High-Quality Methods

High-Quality Methods. Designing Methods Correctly; Cohesion and Coupling. Nikolay Ganchev. Telerik Corporation. www.telerik.com. High-Quality Methods. How to Design and Implement High-Quality Methods? Understanding Cohesion and Coupling. Why do we need Methods?. Jump. start: mov ah,08

phil
Download Presentation

High-Quality Methods

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. High-Quality Methods Designing Methods Correctly; Cohesion and Coupling Nikolay Ganchev Telerik Corporation www.telerik.com

  2. High-Quality Methods How to Design and Implement High-Quality Methods? Understanding Cohesion and Coupling

  3. Why do we need Methods? Jump start: mov ah,08 int 21h mov bl,al JMP output mov ah,01 int 21h output: mov dl,"(" mov ah,02 int 21h mov dl,bl int 21h mov dl,")" int 21h exit: mov ah,4ch mov al,00 int 21h Imagine a long program consisting of instructions and jumps not organized in any structural way

  4. Why We Need Methods? • Methods are important in software development • Reduce complexity • Divide and conquer: complex problems are split into composition of several simple • Improve code readability • Small methods with good method names make the code self-documenting • Avoid duplicating code • Duplicating code is hard to maintain

  5. Why We Need Methods? (2) • Methods simplify software development • Hide implementation details • Complex logic is encapsulated and hidden behind a simple interface • Algorithms and data structures are hidden and can be transparently replaced later • Increase the level of abstraction • Methods address the business problem, not the technical implementation: Bank.accounts[customer].deposit(500);

  6. Using Methods: Fundamentals • Fundamental principle of correct method usage: • Methods should do exactly what their names say • Nothing less • Nothing more • In case of incorrect input or incorrect preconditions, an error should be indicated A method should do what its name says or should indicate an error. Any other behaviour is incorrect!

  7. Good Methods – Examples long Sum(int[] elements) { long sum = 0; foreach (int element in elements) { sum = sum + element; } return sum; } double CalcTriangleArea(double a, double b, double c) { if (a <= 0 || b <= 0 || c <= 0) { throw new ArgumentException("Sides should be positive."); } double s = (a + b + c) / 2; double area = Math.Sqrt(s * (s - a) * (s - b) * (s - c)); return area; }

  8. Indicating an Error • Method errors are often not indicated in real code as done correctly in the example of the previous slide: internal object getValue(string propertyName) { PropertyDescriptor descriptor = propertyDescriptors[propertyName]; return descriptor.GetDataBoundValue(); }

  9. Indicating an Error (2) • If the property name does not exist • A null reference exception will be thrown (implicitly) • It is not meaningful • Use the correct exception instead: internal object getValue(string propertyName) { PropertyDescriptor descriptor = propertyDescriptors[propertyName]; if (descriptor == null) { throw new ArgumentException (“Property name: “ + propertyName + “ does not exists!”); } return descriptor.GetDataBoundValue(); }

  10. Symptoms of Wrong Methods • Method that does something different than its name, is wrong for at least one of these reasons: • The method sometimes returns incorrect result  bug • The method returns incorrect output when its input is incorrect or unusual  low quality • Acceptable for private methods only • The method does too many things  bad cohesion • The method has side effects  spaghetti code • Method returns strange value when an error condition happens  it should indicate the error

  11. Wrong Methods – Examples long Sum(int[] elements) { long sum = 0; for (int i = 0; i < elements.Length; i++) { sum = sum + elements[i]; elements[i] = 0; // Hidden side effect } return sum; } double CalcTriangleArea(double a, double b, double c) { if (a <= 0 || b <= 0 || c <= 0) { return 0; // Incorrect result } double s = (a + b + c) / 2; double area = Math.Sqrt(s * (s - a) * (s - b) * (s - c)); return area; }

  12. Strong Cohesion • Methods should have strong cohesion • Should address single task and address it well • Should have clear intent • Methods that address several tasks in the same time are hard to be named • String cohesion is used in engineering • In computer hardware any PC component solves a single task • E.g. hard disk performs a single task – storage

  13. Acceptable Types of Cohesion • Functional cohesion(independent function) • Method performs certain well-defined calculation and returns a single result • The entire input is passed through parameters and the entire output is returned as result • No external dependencies or side effects • Examples: Math.Sqrt(value)  square root Char.IsLetterOrDigit(ch) String.Substring(str, startIndex, length)

  14. Acceptable Types of Cohesion (2) • Sequential cohesion(algorithm) • Method performs certain sequence of operations to perform a single task and achieve certain result • It encapsulates an algorithm • Example: • Connect to mail server • Send message headers • Send message body • Disconnect from the server SendEmail(recipient, subject, body)

  15. Acceptable Types of Cohesion (3) • Communicational cohesion(common data) • A set of operations used to process certain data and produce a result • Example: • Retrieve input data from database • Perform internal calculations over retrieved data • Build the report • Format the report as Excel worksheet • Display the Excel worksheet on the screen DisplayAnnualExpensesReport(int employeeId)

  16. Acceptable Types of Cohesion (4) • Temporal cohesion (time related activities) • Operations that are generally not related but need to happen in a certain moment • Examples: • Load user settings • Check for updates • Load all invoices from the database • Sequence of actions to handle the event InitializeApplication() ButtonConfirmClick()

  17. Unacceptable Cohesion • Logical cohesion • Performs a different operation depending on an input parameter • Incorrect example: • Can be acceptable in event handlers (e.g. the KeyDown event in Windows Forms) object ReadAll(int operationCode) { if (operationCode == 1) … // Read person name else if (operationCode == 2) … // Read address else if (operationCode == 3) … // Read date … }

  18. Unacceptable Cohesion • Coincidental cohesion (spaghetti) • Not related (random) operations are grouped in a method for unclear reason • Incorrect example: • Prepares annual incomes report for given customer • Sorts an array of integers in increasing order • Calculates the square root of given number • Converts given MP3 file into WMA format • Sends email to given customer HandleStuff(customerId, int[], ref sqrtValue, mp3FileName, emailAddress)

  19. Loose Coupling • What is loose coupling? • Minimal dependences of the method on the other parts of the source code • Minimal dependences on the class members or external classes and their members • No side effects • If the coupling is loose, we can easily reuse a method or group of methods in a new project • Tight coupling  spaghetti code

  20. Loose Coupling (2) • The ideal coupling • A methods depends only on its parameters • Does not have any other input or output • Example: Math.Sqrt() • Real world • Complex software cannot avoid coupling but could make it as loose as possible • Example: complex encryption algorithm performs initialization, encryption, finalization

  21. Coupling – Example • Intentionally increased coupling for more flexibility (.NET cryptography API): byte[] EncryptAES(byte[] inputData, byte[] secretKey) { Rijndael cryptoAlg = new RijndaelManaged(); cryptoAlg.Key = secretKey; cryptoAlg.GenerateIV(); MemoryStream destStream = new MemoryStream(); CryptoStream csEncryptor = new CryptoStream( destStream, cryptoAlg.CreateEncryptor(), CryptoStreamMode.Write); csEncryptor.Write(inputData, 0, inputData.Length); csEncryptor.FlushFinalBlock(); byte[] encryptedData = destStream.ToArray(); return encryptedData; }

  22. Loose Coupling – Example • To reduce coupling we can make utility classes • Hide the complex logic and provide simple straightforward interface (a.k.a. façade): byte[] EncryptAES(byte[] inputData, byte[] secretKey) { MemoryStream inputStream = new MemoryStream(inputData); MemoryStream outputStream = new MemoryStream(); EncryptionUtils.EncryptAES( inputStream, outputStream, secretKey); byte[] encryptedData = outputStream.ToArray(); return encryptedData; }

  23. Tight Coupling – Example • Passing parameters through class fields • Typical example of tight coupling • Don't do this unless you have a good reason! class Sumator { public int a, b; int Sum() { return a + b; } static void Main() { Sumator sumator = new Sumator() { a = 3, b = 5 }; Console.WriteLine(sumator.Sum()); } }

  24. Tight Coupling in Real Code • Say, we have a large piece of software • We need to update subsystems and the subsystems are not really independent • E.g. a change in filtering affects sorting, etc: class GlobalManager { public void UpdateSorting() {…} public void UpdateFiltering() {…} public void UpdateData() {…} public void UpdateAll () {…} }

  25. Cohesion Problems in Real Code • Say, we have an application consisting of two layers: • Do not use a method for both top-down and bottom-up updates! • Both updates are essentially different, e.g. a RemoveCustomer routine in the DataLayer. Presentation Layer Data Layer

  26. Loose Coupling and OOP • Reducing coupling with OOP techniques • Abstraction • Define a public interface and hide the implementation details • Encapsulation • Make methods and fields private unless they are definitely needed • Define new members as private • Increase visibility as soon as this is needed

  27. Acceptable Coupling • Method is coupled to its parameters • This is the best type of coupling • Method in a class is coupled to some class fields • This coupling is usual, do not worry too much • Method in a class is coupled to static methods, properties or constants in external class • This is normal, usually is not a problem • Method is coupled to external static fields • Avoid this, classes should keep their fields private

  28. Non-Acceptable Coupling • Method in a class is coupled to static fields in external class • Avoid it  consider refactoring • Methods take as input data some fields that could be passed as parameters • Check the intent of the method • Is it designed to process internal class data or is utility method? • Method is defined public without being part of the public class's interface  possible coupling

  29. Methods Parameters • Put most important parameters first • First put the main input parameters • Put non-important optional parameters last • Example: • Incorrect example: void RegisterUser(string username, string password, Date accountExpirationDate, Role[] roles) void RegisterUser(Role[] roles, string password, string username, Date accountExpirationDate,) void RegisterUser(string password, Date accountExpirationDate, Role[] roles, string username)

  30. Methods Parameters (2) • Do not modify the input parameters • Use new variable instead • Incorrect example: • Correct example: bool CheckLogin(string username, string password) { username = username.ToLower(); … // Check the username / password here } bool CheckLogin(string username, string password) { string usernameLowercase = username.ToLower(); … // Check the username / password here }

  31. Methods Parameters (3) • Use parameters consistently • Use the same names and the same order in all methods • Incorrect example: • Output parameters should be put last void EncryptFile(Stream input, Stream output, string key); void DecryptFile(string key, Stream output, Stream input); FindCustomersAndIncomes(Region region, out Customer[] customers, out decimal[] incomes)

  32. Pass Entire Object or Its Fields? • When should we pass an object containing few values and when these values separately? • Sometime we pass an object and use only a single field of it – it is a good practice? • Examples: • Look at the method's level of abstraction • Is it intended to operate with employees of with rates and months?  the first is incorrect CalculateSalary(Employee employee, int months); CalculateSalary(double rate, int months);

  33. How Much Parameters Methods Should Have? • Limit the number of method parameters to 7 (+/-2) • 7 is a magic number in psychology • Human brain cannot process more than 7 (+/-2) things in the same time • If the parameters need to be too much reconsider the method intent  does it have clear intent? • Consider putting few parameters in a new class

  34. Methods Length • How long should a method be? • There is no specific restriction • Avoid methods longer than one screen (30 lines) • Long methods are not always bad • Be sure you have a good reason for their length • Cohesion and coupling are more important than the method length! • Long methods often contain portions that could be extracted as separate methods with good name and clear intent  check this!

  35. Pseudocode • Pseudocode can be helpful in: • Routines design • Routines coding • Code verification • Cleaning up unreachable branches in a routine

  36. Pseudocode (2) • What the routine will abstract i.e. the information a routine will hide? • Routine input parameters • Routine output • Preconditions • Conditions that have to be true before a routine is called • Postconditions • Conditions that have to be true after routine execution

  37. Pseudocode (3) • Why it is better to spend time on design before you start coding: • The functionality may be already available in a library, so you do not need to code at all! • You need to think of the best way to implement the task considering your project requirements • If you fail on writing the code right the first time, you need to know that programmers get emotional to their code

  38. Pseudocode Example Routine that evaluates an aggregate expression for a database column (e.g. Sum, Avg, Min) Parameters: Column Name, Expression Preconditions: Check whether the column exists and throw an argument exception if not If the expression parser cannot parse the expression throw an ExpressionParsingException Routine code: Call the evaluate method on the DataView class and return the resulting value as string

  39. Pseudocode Key Points • Design is iterative in nature • When you spend time on design you find shortcomings that you will otherwise find later • When it will be harder to fix them • Use natural language in writing pseudocode • Pseudocode can be used as comments thus documenting the code

  40. Public Routines in Libraries • Public routines in libraries and system software is hard to change, because customers demand no breaking changes • Two reasons why you need to change a public routine: • New functionality has to be added conflicting with the old features • The name is confusing and makes the usage of the library unintuitive

  41. Routine Deprecation • When deprecating an old routine always include that in the documentation and specify the new routine that has to be used in its place • Example: • Obsolete attribute in .NET Framework [Obsolete("CreateXml is deprecated. Use CreateXmlReader instead.")] public void CreateXml (…) { … }

  42. Inline Routines • Inline routines provide two benefits: • Abstraction • Performance benefit of not creating a new routine on the stack • Some applications (e.g. games) need that optimization • Used for the most frequently used routines • Example: a short routine called 100,000 times • Not all languages support Inline routines

  43. Conclusion • Designing and coding routines is engineering activity • There is no perfect solution • Competing solutions usually demonstrates different trade-offs • The task of the programmer is to evaluate the requirements and choose the most appropriate solution from the available options

  44. High-Quality Methods ? Questions? ? ? ? ? ? ? ? ? ? http://academy.telerik.com

More Related